Periodica Polytechnica Three Dimensional Finite Element

Civil Engineering
Analysis of Deep Excavations
Concave Corners
60(3), pp. 371378, 2016
DOI: 10.3311/PPci.8608 Attila Szepeshzi, Andrs Mahler, Balzs Mczr
Creative Commons Attribution

RESEARCH ARTICLE
Received 24-09-2015, accepted 03-10-2015

Abstract
The behavior of deep excavations retaining structures may
become rather complex near the corners. The vicinity of con-
cave corners is an especially problematic zone, and causes nu-
merous problems in everyday practice. Conventional 2D anal-
yses considering axisymmetric or plane strain conditions can-
not be used in such cases due the spatial nature of the prob-
lem. A set of three dimensional finite element analyses have
been performed aiming better understanding of soil and struc-
ture behavior near excavations concave corners and evaluating
the influence of corner angle. The bending moment arising in
anisotropic diaphragm walls near the concave corners are sum-
marized, and compared to 2D plane strain and axisymmetric
results.
Keywords
deep excavation retaining structures 3D finite-element
analysis concave corner
Attila Szepeshzi
Department of Engineering Geology and Geotechnics Faculty of Civil Engi-
neering, Budapest University of Techology and Economics, H-1521 Budapest,
P.O.B. 91, Hungary
e-mail: szepeshazi.attila@epito.bme.hu
Andrs Mahler
Department of Engineering Geology and Geotechnics Faculty of Civil Engi-
neering, Budapest University of Techology and Economics, H-1521 Budapest,
P.O.B. 91, Hungary
e-mail: mahler.andras@epito.bme.hu
Balzs Mczr sumes plane strain conditions for the middle parts and in some
Department of Engineering Geology and Geotechnics Faculty of Civil Engi-
neering, Budapest University of Techology and Economics, H-1521 Budapest,
P.O.B. 91, Hungary
e-mail: moczar.balasz@epito.bme.hu
1 Introduction
Geotechnical problems are often complex in geometry and
therefore 3D analysis is required for reliable design. 3D finite el-
ement modelling are more widespread in some special geotech-
nical fields, such as tunnelling. Nevertheless special factors,
such as time dependent material models are still only analysed
in 2D [1]. Concave corners of deep excavations are a typical ex-
ample where 3D effects cannot be neglected. Additionally this
type of problem is often faced during the everyday practice. Al-
though commercially available 3D finite element softwares be-
came available in the last decade; everyday geotechnical design
is still mostly restricted to 2D analyses. A commercial FE soft-
ware (Midas GTS) was used to create 3D models of excavations
retained by diaphragm walls having concave corners of differ-
ent angles. A set of analyses has been performed to investigate
the diaphragm wall behaviour near the concave corner of a deep
excavation. This paper summarizes the results of this study.
2 Analysis of retaining walls
The design methods of retaining walls can be classified into
3 groups [2, 3]: simplified calculation of the earth pressures ac-
cording to the assumed movement directions of the wall; calcu-
lations based on beam on elastic springs theory; and finite ele-
ment analysis. The first two approaches are not really capable
of taking into account the non-linearity of soil behaviour or for
complex soil-structure interaction problems. In such situation
finite element analyses are used, at which the structure and the
soil can be modelled using finite elements and their interaction
can be considered by defining interface elements. Such analy-
ses may provide realistic predictions of excavation performance,
when appropriate constitutive model and carefully calibrated in-
put parameters are used [4]. Due to the complexity of the prob-
lem, finite element back analyses are often performed to enable
deeper understanding of the soil behaviour [5].
In case of deep excavations, design practice generally as-
cases axisymmetric conditions for the proximity of the corners.
Such assumption may provide reliable results in case of exca-
vations with long sides and consistent load distributions, but for

Three Dimensional Finite Element Analysis of Deep Excavations Concave Corners 2016 60 3 371
problems with more complex geometry or soil conditions may
lead to unreliable results. Based on the results of 150 FE anal-
yses Finno et al. [6] found that surface settlements obtained by
2D analyses are nearly equal to the result of 3D analyses results
only, when the width of the excavation is greater than six times
the excavation depth.
A detailed study based on 2D and 3D finite element analyses
of a 40 m deep and 35 m wide excavation [7] pointed out that
anisotropy of wall stiffness may also have a significant effect
on the computed results. Different types of retaining walls used
in underground construction, such as pile walls or diaphragm
walls, have much lower horizontal stiffness compared to the
vertical one. Thus, the load bearing occurs mainly by resist-
ing bending moments rotating around the horizontal axis. Fig. 1
shows the comparison of the bending moments rotating around
the horizontal axis obtained by Zdravkovic using different con-
siderations. A quite good correlation of the curves from different
models can be observed. Nevertheless, there is a significant dif-
ference in the upper 15 - 20 m of the wall when isotropic wall
stiffness is considered in 3D model.
Fig. 1. Bending moments obtained from the different models [7]
3 Behaviour of concave corners
Many study aimed to analyse the behaviour of deep exca-
vations retaining structures near the corners [810], but these
studies focus on positive (convex) corners.
An excavation can be interpreted as a steep earth mass re-
tained by a wall structure. Thus the stabilizing actions are
amended with the reaction of the supporting structures. The
sketches in Fig. 2 show a straight part and a concave corner of
an excavation and the potential slip surfaces. In the case of con-
cave corners the volume of sliding wedge is somewhat smaller
than that in the case of an infinite wall (i.e. plane strain), but the
difference is not significant. On the other hand the area of the
sliding surface decreases dramatically especially in the zones of
high stress level, where significant shear resistance can develop.
372 Period. Polytech. Civil Eng.
Thus a wall retaining such distribution of earth mass must resist
higher destabilizing actions compared to straight wall sections
of the same height. Furthermore, the concave corner in an exca-
vation must usually be constructed because there is an existing
structure which represents a high surface load. Thus we can
state that the destabilizing actions are in a similar range in both
cases.
Fig. 2. The analysis of a sliding wedge
The possible solutions are different in case of different retain-
ing structure types. The most frequently used types; pile wall
and diaphragm wall are discussed here.
In case of non-contiguous pile walls, the single piles are con-
nected by a capping beam and possibly by some braces for the
struts or anchors. Therefore, the horizontal stiffness of the wall
is usually quite low and the wall behaviour can be interpreted
by the resistance of single piles. The major difference in case
of a pile near the concave corner, compared to a straight wall
section, is that the main direction of the bending moments act-
ing on the piles is not perpendicular to the wall plane. However
the construction experiences show that cracking of piles, larger
displacements and problems with the waterproofing occur fre-
quently. Auxiliary solutions are usually necessary, such as in-
stallation of extra props or grouting in between the piles.
The case of diaphragm walls is more complex and special,
and the details of construction may cause non-negligible differ-
ences in excavation performance [11, 12]. The connection of
the adjoining panels cannot resist tension and bending moments
and its shear resistance in the direction perpendicular to the wall
plane is also limited. Due to wall deformations caused by earth
pressure, the panels joining along the edge of the concave cor-
ner would split up without modification of the usual joint de-
sign. Thus a possible construction solution can be to link the
reinforcement of these panels and to form a rigid corner. De-
spite this technique, occasional cracking of the wall cannot be
avoided. Therefore similar auxiliary methods can be utilized
as mentioned previously. With respect to the special geometry
and structural distribution, the application of traditional design
procedures is inaccurate, the usual design of concave corners is
based on previous experiences, and generally higher factors of
safety are applied.
A short remark must be made for the very rapid technologi-
Attila Szepeshzi, Andrs Mahler, Balzs Mczr
cal development of deep excavation construction techniques: so-
called deep mixing walls made of soil-binder mixtures with steel
profiles are used in practice for shallower excavations. The ap-
plication of such innovative construction techniques can be ex-
pected for deeper excavations, as well. Therefore future research
shall focus on realistic modelling of concave corners of such
wall structures which can be totally different from diaphragm-
wall behaviour analysed in present paper. Prediction of deep
mixing materials characteristics are a present-day research topic
[13] making the issue even more complicated.
As a conclusion it can be stated, that no special design tech-
nique exists to find an economic solution for walls retaining con-
cave corners. Instead application of higher safety factors or aux-
iliary construction procedures is used.
4 Finite element analysis
4.1 The 3D analysis series
Three dimensional FE analyses were carried out to analyse
the internal forces and displacements of the wall retaining the
concave corners having different corner angles. Fig. 3 shows the
four different layouts created. The wall sections around the cor-
ner are 6 m long while the adjoining wall sections are 7, 8, 9 and
10 m long respectively. The total height of the wall is 18 m, the
depth of the excavation is 8 m thus the embedment depth is 10 m.
To avoid the possible negative effect of larger displacements on
the accuracy of the calculations, the excavation is carried out in
four steps.
As it was discussed previously the application of a correct
stiffness distribution and realistic modelling of structural joints
can strongly influence the results. During the analyses 80 cm
thick and 3 m wide cantilever diaphragm wall panels were mod-
elled with isotropic plate elements. The joints of the adjacent
panels are hinged thus rotation around the vertical axis is al-
lowed. The only exception is the joint in the edge of the concave
corner where fixed connections have been defined. To obtain
a correct soil-structure interaction, interface elements were de-
fined among the solid elements of the soil and the plate elements
of the wall.
Fig. 3. The layout of the excavations
Three Dimensional Finite Element Analysis of Deep Excavations Concave Corners
Modelling soil behaviour generally requires the use of ad-
vanced soil models, however their application increases the cal-
culation time significantly, especially in case of three dimen-
sional models [14]. Therefore, a linear elastic perfectly plastic
(Mohr-Coulomb) model was used for the calculations. In or-
der to simplify the problem and the evaluation, a single sand
layer was considered, the properties defined for the diaphragm
wall, for the soil and for the interface elements are summarized
in Table 1 and Table 2.
The finite element model of the diaphragm wall were built of
0.5 0.5 m size rectangle shaped elements. The solid elements
of the soil inside the excavation were fitted to this size while
outside the excavation their sizes were increased linearly with
increasing distance from the wall. Thus the size of the solid ele-
ments at the boundary is 5 m. The finite element mesh is shown
in Fig. 4, the number of elements and nodes are summarized in
Table 3.
The diaphragm wall installation and the soil excavation were
defined in six construction stages, considering four excavation
steps. A distributed surface load of 10 kPa was applied to avoid
the possible numerical problems, which may arise from the de-
tachment of the soil and wall elements in shallow regions due to
the small soil cover.
Fig. 4. The finite element mesh
4.2 The results of the model with a concave corner of right
angle
Fig. 5 presents the total displacements (DXYZ) of the model
in hundredfold magnification. The panel joints behave as ex-
pected: the hinged joints are rotated while the edge of the con-
cave corner behaves rigidly. The red shade of the bottom means
about 4 cm heave. This unrealistic result is caused by the con-
sidered linear elastic soil model, which does not differentiate
between primary loading and unloading-reloading. Such errors
can be avoided by using more sophisticated soil models. A rela-
tively large embedment depth was used in the analysis to mini-
mize the effect of base heave on the retaining wall deformations.
The same results from a different point of view can be seen
on Fig. 6. This shows the total displacements of the half model
intersected in its symmetry plane. The wall panels and the soil
mass inside the walls are hidden. The shades of the concave
2016 60 3 373
 Tab. 1. Material properties

Youngs Unit Internal

Material Poissons Cohesion
Type modulus E weight friction K0 [-]
model ratio [-] c [kPa]
[kPa] [kN/m3 ] angle []
Wall Elastic 35000000 0.2 24 - - -
Mohr
Soil 30000 0.3 20 1 30 0.5
Coulomb

Tab. 2. Interface properties

Normal Stiffness Shear Stiffness Internal friction

Element Cohesion c [kPa]
Modulus Kn [kN/m3 ] Modulus Kt [kN/m3 ] angle []
Interface 6000000 60000 15 1

corner shows similar slipping earth mass as it was described on

Fig. 2. The displacements are quite low, the red shades show the
highest values of about 1.4 cm.
The bending moments rotating around the horizontal axis are
shown on Fig. 7. The positive values (green, yellow and red
shades) illustrate the case when the outer side of the wall is in
tension while the negative values (blue shades) mean the oppo-
site direction of the bending moments. Fig. 7 only shows one
half of the wall structure to obtain better visibility. Contrary to
the expectations the maximum values of these bending moments
are much lower around the concave corner (~ 40 kNm) than at Fig. 6. Total displacements (DXYZ) at the final stage

the longer, straight wall section (~ 135 kNm). Although the lin-
ear plastic soil model may overestimate the soil expansion at the
excavation base, thus underestimate the positive moments, the
tendencies are clear; the moments rotating around the horizon-
tal axis do not seem to be the major problem in case of concave
corners.
from the plane strain analyses are compared with the bending
moments arising at the mid span of the 15 metre long, straight
wall section. The results of the axisymmetric calculations are
compared with the bending moments obtained from the proxim-
ity of the convex corner in the same straight wall section.

Fig. 7. Bending moments rotating around the horizontal axis at the final
stage

The 3D model in the proximity of the convex corner and the

Fig. 5. Total displacements (DXYZ) at the final stage
axisymmetric models result in bending moments and deforma-
tions varying over similar range, but showing different tenden-
Two dimensional finite element analyses have also been per-
cies. The two curves are nearly the opposite of each other, im-
formed to enable a better evaluation of the results of 3D analy-
plying that analysis of an axisymmetric silo-like structure could
ses. Plane strain and axisymmetric models were created using
provide results that are more reliable. However, due to the ex-
the same soil and interface properties
tremely small deformations and bending moments here, the im-
Fig. 8 shows the comparison of the computed bending mo-
portance of this topic is negligible.
ments of the 2D and 3D models. The inner forces resulting

374 Period. Polytech. Civil Eng. Attila Szepeshzi, Andrs Mahler, Balzs Mczr
 Tab. 3. Mesh characteristics

Element/Node 72 90 112 142

Wall (2D) 4464 4320 4176 4032
Number of elements
Soil (3D) 11391 11020 10824 10706
Number of nodes 33141 32095 31366 30778

Fig. 8. Comparison of the results obtained by 2D and 3D analysis

The plane strain analyses showed ~ 7.5 cm horizontal dis-

placement at the top of the wall while the 3D model resulted
~ 1.4 cm only. Such significant difference appears in the bending
moment results too: neither the maximum value nor the shape
of the curve show similarity in the case of the two assumptions.
The 3D model changes sign around the bottom part of the exca-
vation while the 2D model presents that the outer side of the wall
is in tension all along. This implies that simplifications, which
are generally used for 2D calculations (such as neglecting arch-
ing effect, anisotropic wall behaviour) are not acceptable in case
of short walls and their application leads to significant overdi- Fig. 9. Bending moments rotating around the vertical axis at the final stage
mensioning.
The other important question related to the 3D analyses re-
layouts. Here the bending moments arose in section number 3
sults is the distribution of bending moments rotating around the
are shown only as the highest values are experienced here. Ac-
vertical axis. Fig. 9 shows these results. The shades can be inter-
cording to the diagram on Fig. 11, the bending moments increase
preted as it was described in case Fig. 7. These results reveal the
with the increasing corner angles. With respect to the previous
tendency that may lead to problems during construction. The
figures it is reasonable due to the fact that higher angle means
bending moments around the concave corner are significantly
higher mass of retained earth.
higher than the values at straight wall section divided by hinged
The bending moments rotating around the vertical axis were
joints. The extreme maximum of bending moments arises at
compared in 8 different horizontal segments marked in Fig. 12.
the edge of the concave corner presented by the blue shades on
The origin of z axis is at the excavation bottom, sections with
Fig. 9. Such moments lead to the structural problems that are
positive z coordinates are located above this level, negative val-
experienced in the everyday construction practice.
ues below. Section number 1 (Fig. 13), 4 (Fig. 14) and 6
(Fig. 15) are presented in this paper only.
4.3 Effect of corner angle
Hinged joints are defined along the both sides of the second
The previous figures showed that in this case the bending mo-
wall panel retaining the concave corner. Accordingly, the bend-
ments rotating around the horizontal axis are not as high as the
ing moments rotating around the vertical axis are zero at the
bending moments rotating around the vertical axis. Nevertheless
connections as it can be noticed in Fig. 13, 14 and 15.
a short examination must be made. Fig. 10 shows four vertical
However, the absolute value of the bending moment in the
sections where the bending moments were analysed in all the

Three Dimensional Finite Element Analysis of Deep Excavations Concave Corners 2016 60 3 375
 Fig. 12. The examined horizontal sections of the wall
Fig. 10. The analysed vertical sections of the wall retaining the concave cor-
ner

Fig. 13. The bending moments in section number 1 (z = 7,75 m) at the final
stage
Fig. 11. The bending moments arose in section 3 at the final stage

In the embedded section the bending moments tend to have

middle of this panel changes with the angle of the corner. The
positive values again but extreme values cannot be found there.
models with the angle of 112and 142show approximately
Thus it can be stated that the lower angles of the concave corner
30 kNm in section number 1 and their magnitude changes only
cause higher bending moments.
slightly in deeper sections. Beneath the bottom of the excava-
Fig. 16 illustrates these tendencies in a different way a section
tion, these forces disappear nearly linearly. In the case of models
was taken at the corner and moments rotating around the verti-
with the angle of 72and 90the maximal bending moments are
cal axis are plotted against depth as well. The curves moves to
also around 30 kNm but their value increase with depth. The
the negative part as the corner angle decreases and the negative
absolute maximum is given in section 4 where the model with
extreme increases significantly. The same tendency results in
the right angle shows 50 kNm and the model with the angle of
decreasing positive extreme value.
72shows 60 kNm. Beneath the bottom of the excavation these
values disappear linearly. It can be seen that the bending mo-
5 Conclusions
ments are higher with lower angle of the concave corner.
A diaphragm wall retaining an 8 m deep excavation has been
The wall panels adjoining in the edge of the corner are more
analysed by 2D and 3D finite element models. Special atten-
of practitioners interest. The bending moments change sign in
tion has been paid to the behaviour of concave corners of dif-
all the models except the model with the angle of 72. The bend-
ferent angles. Joints have been defined at the connection of the
ing moments tend to be negative with increasing depth. The
diaphragm wall panels to take into account their anisotropic be-
models provide their maximum value of bending moment in
haviour and rigid connection has been chosen for the concave
section 4, except the model with the angle of 142. This ten-
corner.
dency fits the previous expectations as section 4 is a bit above
The computed bending moments rotating around the horizon-
the bottom of the excavation. This maximum values are 86,
tal axis are significantly smaller in case of 3D analysis than in
135 and 187 kNm in the models with the angle of 112, 90and
the case of 2D plane strain model. This is in good agreement
72respectively.

376 Period. Polytech. Civil Eng. Attila Szepeshzi, Andrs Mahler, Balzs Mczr

Fig. 14. The bending moments at the final stage in section 4 (z = 1,75 m)
Fig. 15. The bending moments at the final stage in section 6 (z = -3,25 m)
Fig. 16. Bending moments rotating around vertical axis vs. depth along the
edge of the concave corner
Three Dimensional Finite Element Analysis of Deep Excavations Concave Corners
with earlier experiences, considering that the chosen excava-
tion width is less than 20 m, therefore the arching effect and
anisotropic wall behaviour causes significant reduction in bend-
ing moments. It also reconfirmed that this internal force is not
the cause of the problem that is generally experienced at concave
corners.
Results of bending moments rotating around the vertical axis
revealed the basic problem of concave corners. They have
shown that the moments increase significantly in the vicinity of
the concave corners, and they can be as large as or even larger
than the moments rotating around the horizontal axis. As this
component cannot be computed using 2D analyses the dimen-
sioning of this part is still a problematic point of design in many
cases, and it is handled with a lot of empiricism. The critical
parts of the structure can be easily located, however the magni-
tude of the bending moments is complicated to estimate. The
magnitude and even the sign of this moment is strongly influ-
enced by the depth; the extreme moments, that causes tension
on the excavation side, developed around the bottom third of the
excavation depth (z = 2.5 - 3.0 m). This area has been reported
the critical zone in case of similar excavations earlier by con-
tractors too.
A clear tendency has been observed regarding the corner an-
gle as well: the larger the corner angle, the bending moments
move to the positive part and in the cases of smaller angles neg-
ative bending moments arise (Fig. 16). Thus significant bending
moments can develop in case of sharp corner angles.
The analysis of this simplified problem helps to understand
the nature of the basic characteristics of soil and structure be-
haviour near concave corners. Recent results helped in locating
the critical parts of the structure and find the critical internal
force components. However, further work is needed to give rec-
ommendation on estimating the value of the bending moments
without 3D finite element analyses.
References
1 Seifabad M C, Sadrnejad S A, Ebrahimi D, A study of swelling behaviour
in a tunnel using finite element methods, Periodica Polytechnica Civil Engi-
neering, 59(2), (2015), 103107, DOI 10.3311/PPci.7538.
2 Yu-Ou Ch, Deep excavation. Theory and Practice, Taylor and Francis; Lon-
don, United Kingdom, 2006.
3 Ray R, Scharle P, Szepeshzi R, Numerical modelling in the geotechnical
design practice, In: Proceedings of 14th Danube-European Conference on
Geotechnical Engineering, 2010.
4 Nikolinakou M A, Whittle A J, Savidis S, Schran U, Prediction and inter-
pretation of the performance of a deep excavation in Berlin sand, ASCE Jour-
nal of Geotechnical and Geoenvironmental Engineering, 137(11), (2011),
10471061.
5 Hash Y M A, Song H, Osouli A, Three-dimensional inverse analyses of
a deep excavation in Chicago clays, International Journal for Numerical
and Analytical Methods in Geomechanics, 35(9), (2010), 10591075, DOI
10.1002/nag.949.
6 Finno R J, Blackburn J T, Roboski J F, Three-"Dimensional Effects for
Supported Excavations in Clay, Journal of Geotechnical and Geoenviron-
mental Engineering, 133(1), (2007), 3036.
2016 60 3 377
 7 Zdravkovic L, Modelling deep excavations in 3D analysis in Deep Excava-
tions and Retaining Structures, In: A. Mahler LN (ed.), ISSMGE Hungarian
National Committee; Budapest, Hungary, 2009.
8 Lin D G, Woo S M, Three dimensional analyses of deep excavation in Taipei
101 construction project, Journal of GeoEngineering, 2(1), (2007), 2941.
9 Wu C H, Ou C Y, Tung N, Corner effects in deep excavations-establishment
of a forecast model for taipei basin T2 zone, Journal of Marine Science and
Technology, 18(1), (2010), 111.
10 Tan Y, Wei B, Diao Y, Zhou X, Spatial Corner Effects of Long and Narrow
Multipropped Deep Excavations in Shanghai Soft Clay, Journal of Perfor-
mance of Constructed Facilities, 28(4), (2014).
11 Zhao B D, Zhang L L, Jeng D S, Wang J H, Chen J J, Inverse Analysis of
Deep Excavation Using Differential Evolution Algorithm, International Jour-
nal for Numerical and Analytical Methods in Geomechanics, (2014), DOI
10.1002/nag.2287.
12 Burlon S, Mroueh H, Shahrour I, Influence of diaphragm wall installa-
tion on the numerical analysis of deep excavation, International Journal for
Numerical and Analytical Methods in Geomechanics, 37(11), (2013), 1670
1684, DOI 10.1002/nag.2159.
13 Arasan S, Isik F, Akbulut R K, Zaimoglu A S, Nasirpur O, Rapid sta-
bilization of sands with deep mixing method using polyester, Periodica Poly-
technica Civil Engineering, 59(3), (2015), 405411, DOI 10.3311/PPci.7956.
14 Potts D M, Zdravkovic L, Finite element analysis in geotechnical engineer-
ing theory, Thomas Telford; London, United Kingdomn, 1999.

378 Period. Polytech. Civil Eng. Attila Szepeshzi, Andrs Mahler, Balzs Mczr